Cleanup _get_enum_from_fn implementation in _api.py

[GdoongMathew]

🚀 The feature

In the current _get_enum_from_fn implementation, there are two things that can be refined.

1. https://github.com/pytorch/vision/blob/main/torchvision/models/_api.py#L167-L171, signature function was called twice, but I think it only needs one call.

def _get_enum_from_fn(fn: Callable) -> Type[WeightsEnum]:
    sig = signature(fn)
    if "weights" not in sig.parameters:
        raise ValueError("The method is missing the 'weights' argument.")

    ann = sig.parameters["weights"].annotation    # <-- here

2. resolve the TODO at line https://github.com/pytorch/vision/blob/main/torchvision/models/_api.py#L177, since torch already removed its support from py3.8, there should be no risk making that modification.

Motivation, pitch

Just saw this and think it'd be helpful?

Alternatives

No response

Additional context

No response

[NicolasHug]

Thanks for the report @GdoongMathew . I agree with you that this could be cleaned up, however I don't think it would lead to a significant improvement - that part of the code-base "just works" and we don't really need to touch it as it's not a hot path.

If you'd like to submit a PR though, I'll make sure to review it.